Retrieval-Augmented Generation in Biomedicine: A Survey of Technologies, Datasets, and Clinical Applications

TL;DR

This survey reviews retrieval-augmented generation (RAG) in biomedicine, discussing its architectures, challenges like latency and privacy, and future directions for clinical applications and multimodal integration.

Contribution

It provides a comprehensive classification, formalizes the biomedical RAG trilemma, and analyzes recent advancements and challenges in deploying RAG for clinical use.

Findings

Classifies biomedical RAG systems into naive, advanced, and modular paradigms.

Identifies trade-offs between reasoning, latency, and privacy in RAG deployment.

Highlights the potential and risks of agentic workflows in diagnostics.

Abstract

Large language models (LLMs) in biomedicine face a fundamental conflict between static parameter knowledge and the dynamic nature of clinical evidence. Retrieval-Augmented Generation (RAG) addresses this by grounding generation in external data, yet it introduces new complexities in latency and architecture. This survey synthesizes the biomedical RAG landscape (2020-2025), classifying systems into naive, advanced, and modular paradigms. Beyond a technological taxonomy, we formalize the biomedical RAG trilemma, identifying the inherent trade-offs between reasoning depth, inference latency, and data privacy that constrain current clinical deployment. We analyze how recent agentic workflows enhance diagnostic reasoning but risk prohibitive latency, and how privacy constraints dictate the choice between powerful cloud-based models and local deployment. Finally, we outline the alignment gap in multimodal RAG and propose future directions for self-correcting, verifiable clinical agents.